Mutations in Both KRAS and BRAF May Contribute to the Methylator Phenotype in Colon Cancer

GASTROENTEROLOGY 2008;134:1950 –1960

BASIC–ALIMENTARY TRACT Mutations in Both KRAS and BRAF May Contribute to the Methylator Phenotype in Colon Cancer TAKESHI NAGASAKA,* MINORU KOI,* MATTHIAS KLOOR,‡ JOHANNES GEBERT,‡ ALEX VILKIN,* NAOSHI NISHIDA,* SUNG KWAN SHIN,* HIROMI SASAMOTO,§ NORIAKI TANAKA,§ NAGAHIDE MATSUBARA,§ C. RICHARD BOLAND,* and AJAY GOEL* *Division of Gastroenterology, Department of Internal Medicine, and Charles A. Sammons Cancer Center and Baylor Research Institute, Baylor University Medical Center, Dallas, Texas; ‡Institute of Molecular Pathology, University of Heidelberg, Heidelberg, Germany; and §Department of Gastroenterological Surgery and Surgical Oncology, Okayama University Graduate School of Medicine Dentistry and Pharmaceutical Sciences, Okayama, Japan

BASIC– ALIMENTARY TRACT

Background & Aims: Colorectal cancers (CRCs) with the CpG island methylator phenotype (CIMP) often associate with epigenetic silencing of hMLH1 and an activating mutation in the BRAF gene. However, the current CIMP criteria are ambiguous and often result in an underestimation of CIMP frequencies in CRCs. Because BRAF and KRAS belong to same signaling pathway, we hypothesized that not only mutations in BRAF but mutant KRAS may also associate with CIMP in CRC. Methods: We determined the methylation status in a panel of 14 markers (7 canonical CIMP-related loci and 7 new loci), microsatellite instability status, and BRAF/KRAS mutations in a collection of 487 colorectal tissues that included both sporadic and Lynch syndrome patients. Results: Methylation analysis of 7 CIMP-related markers revealed that the mean number of methylated loci was highest in BRAF-mutated CRCs (3.6) vs KRAS-mutated (1.2, P < .0001) or BRAF/KRAS wild-type tumors (0.7, P < .0001). However, analyses with 7 additional markers showed that the mean number of methylated loci in BRAF mutant tumors (4.4) was the same as in KRAS mutant CRCs (4.3, P ⴝ .8610). Although sporadic microsatellite instability high tumors had the highest average number of methylated markers (8.4), surprisingly, Lynch syndrome CRCs also demonstrated frequent methylation (5.1). Conclusions: CIMP in CRC may result from activating mutations in either BRAF or KRAS, and the inclusion of additional methylation markers that correlate with mutant KRAS may help clarify CIMP in future studies. Additionally, aberrant DNA methylation is a common event not only in sporadic CRC but also in Lynch syndrome CRCs.

A

berrant promoter hypermethylation associated with transcriptional silencing of multiple tumor suppressor genes has been proposed as a mechanistic component in the evolution of several human cancers.1 Tumors with a critical degree of aberrant methylation

have the CpG island methylator phenotype (CIMP), which was initially described in colorectal cancer (CRC).2 CIMP, microsatellite instability (MSI), and chromosomal instability (CIN) constitute the 3 major mechanisms of genomic or epigenetic instability in CRC.3,4 Since the initial description of CIMP, its role in carcinogenesis has been controversial.3,5–7 Experimental evidence has consistently supported the presence of CIMP in a subset of CRCs,3,4,8,9 and it has been found that CIMP significantly correlates with a V600E mutation in the BRAF gene.10,11 However, some degree of promoter methylation can be found in normal tissues and virtually all CRCs, so some have argued that CIMP does not constitute a unique pathogenetic pathway.7 Because of the controversy surrounding CIMP, many laboratories have focused on making a case that CIMP constitutes a discrete group of CRCs through the identification of an optimized panel of CIMP-specific markers.11–13 Two panels of CIMP-specific markers have been proposed, including those identified by the originators of the concept,12,13 and a second panel recently reported after an extensive evaluation of a very large group of candidate markers.11 Both panels are highly specific for the identification of CIMP CRCs with BRAF V600E mutations.11 However, there is a lack of consensus on the frequency of CIMP in CRC because of multiple definitions of CIMP (CIMP⫹ and CIMP⫺, CIMP-high and low, and others), the use of different sets of methylation markers, and differences in diagnostic methodologies. However, the greatest challenge has been to use consistent criteria for CIMP Abbreviations used in this paper: CIMP, CpG island methylator phenotype; C-N, normal colorectal tissue from a patient with colorectal cancer; CRC, colorectal cancer; MSI, microsatellite instability; MSI-H, MSI-high; MSS, microsatellite stable; N-N, colorectal tissue from a normal patient (without colorectal neoplasia). © 2008 by the AGA Institute 0016-5085/08/$34.00 doi:10.1053/j.gastro.2008.02.094

analyses to categorize a cancer positive or negative for this phenotype and to find a consensus for the definition of CIMP.8,9,14,15 Nearly all of the approximately 15% of sporadic MSI CRCs come from a CIMP background, caused by the epigenetic silencing of hMLH1; however, additional CRCs that are microsatellite stable (MSS) may also be labeled as CIMP depending on the diagnostic criteria used.4,8,9 It has been proposed that CIN and CIMP represent 2 major mechanisms of genomic (or epigenetic) instability in CRC and that perhaps up to 50% of CRCs might be characterized as having CIMP.4 Thus, it is important to have a more complete understanding of CIMP and to reach a consensus on whether this constitutes a unique and unified group of tumors that may evolve through a common pathway. CIMP determinations using CIMP-related markers have consistently identified clusters of CRCs with MSI and V600E BRAF mutations but rarely KRAS mutations.8,9,14,16 However, when additional methylation loci have been investigated, additional subsets of CRCs have been identified with extensive methylation; these tumors are non-MSI or MSS and are associated with mutations in the KRAS gene.4,17,18 Although these tumors should be considered CIMP positive, they have been typically categorized as nonCIMP because, by definition, analyses using the canonical CIMP criteria considered those methylation events that associated with mutant BRAF. Defining the markers in this fashion raises a possible problem of circular reasoning. BRAF and KRAS gene products function in the same serial signaling pathway, and activating mutations in these genes occur mutually exclusively.10,19 Interestingly, prior to the more recent discovery of V600E BRAF mutations, KRAS mutations were proposed as the possible cause of aberrant methylation. It has been shown that fibroblasts transformed by fos or ras experience up-regulation of DNA methyltransferase expression and consequent global hypermethylation.20 In light of these considerations, we hypothesized that CIMP in CRC may not be exclusively the result of mutations in BRAF but might be a more general consequence of the up-regulation of the RAS-RAF pathway. In the present study, we investigated the relationship between the activating mutations of the BRAF/KRAS genes and widespread hypermethylation using a broad panel of methylation markers that included both the canonical CIMP-related markers and the additional methylationrelated loci. We included in our investigation a group of Lynch syndrome-related CRCs, in which there is MSI, but no mutations in BRAF. We found that methylation of this broad panel of methylation markers segregates not just with BRAF but also with KRAS mutations. Our data provide novel evidence for a potential role of KRAS mutations in the evolution of aberrant methylation and propose a broader panel of methylation markers that

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

1951

may improve our current understanding for the molecular basis of CIMP.

Materials and Methods Colorectal Tissue Specimens We collected 487 colorectal tissues, including 243 cancers, 208 corresponding normal colonic mucosal specimens from patients with CRC (C-N), and 36 normal epithelial tissues from patients without any evidence of neoplasia at colonoscopy (N-N) from Okayama University Hospital, Okayama, Japan, and Heidelberg University, Heidelberg, Germany. Of the 243 CRCs, 21 cancers were from Lynch syndrome patients and had documented germ-line mutations and associated loss of protein expression of the MMR genes: hMSH2, hMLH1, hPMS2, or hMSH6; the remaining 222 were sporadic CRCs, in which patients did not have a strong family history of CRC and had no evidence for polyposis or inflammatory bowel disease. Among 222 sporadic CRCs, 184 cases have been previously analyzed for CIMP.17 Normal mucosal tissues were sampled from distant surgical margins of the CRC resections. All patients provided written informed consent, and the studies were approved by institutional review boards of all institutions involved.

Sodium Bisulfite Modification and Combined Bisulfite Restriction Assays Genomic DNA obtained from colorectal tissues was bisulfite modified to convert all unmethylated cytosine residues to uracils. Bisulfite polymerase chain reactions (PCR) were carried out, and restriction endonuclease digestion was performed on PCR products for methylation analyses of 14 markers (Table 1).21 Seven methylation markers were canonical CIMP-related loci, including the hMLH1-5= region, hMLH1-3= region, p16INK4a, p14ARF, MINT1, MINT2, and MINT31.4,5,16 We then analyzed methylation frequencies at promoter regions of 7 additional tumor suppressor genes that have been reported to be frequently methylated in a cancer-specific manner in CRC (SFRP2, RASSF2A, MGMT, Reprimo, 3OST2, HPP1, and APC).22–29 The digested DNA was separated on 3% agarose gels and stained with ethidium bromide. The quantitative methylation levels (ratios of methylated to unmethylated DNA) were determined from the relative intensities of cleaved and noncleaved PCR products. A marker was considered methylation positive if it showed ⱖ5% methylation density and methylation negative if it had ⬍5%, in accordance with canonical CIMP criteria.12

Microsatellite Analysis Microsatellite analysis of each tumor tissue was determined using the National Cancer Institute workshop panel of recommended markers.30 Tumors showing a shift in at least 1 mononucleotide marker and 1 other marker were classified as microsatellite instability high

BASIC– ALIMENTARY TRACT

June 2008

1952

NAGASAKA ET AL

GASTROENTEROLOGY Vol. 134, No. 7

Table 1. Primer Sequences, Restriction Endonuclease, and PCR conditions for COBRA Locus

Primer sequence

hMLH1-5=

F: 5=YGGGTAAGTYGTTTTGAYGTAGA3= R: 5=ATACCTAATCTATCRCCRCCTCATC3= F: 5=GGGAGGGAYGAAGAGATTTAGT3= R: 5=ACCTTCAACCAATCACCTCAAT3= F: 5=GGTTTTGGYGAGGGTTGTTT3= R: 5=ACCCTATCCCTCAAATCCTCTAAAA3= F: 5=TTTYGGGGYGGAGATGGGT3= R: 5=ATCACCAAAAACCTACRCACCATATTC3= F: 5=GGGTTGGAGAGTAGGGGAGTT3= R: 5=CCATCTAAAATTACCTCRATAACTTA3= F: 5=YGTTATGATTTTTTTGTTTAGTTAAT3= R: 5=TACACCAACTACCCAACTACCTC3= F: 5=GAYGGYGTAGTAGTTATTTTGTT3= R: 5=CATCACCACCCCTCACTTTAC3= F: 5=GGTTGTTAGTTTTTYGGGGTTT3= R: 5=AACCAAAACCCTACAACATCRT3= F: 5=TTGGGGAGGGTTTGATAGTTT3= R: 5=CRCACCCTACRCCCCTCTAAAA3= F: 5=GTTTTTAGAAYGTTTTGYGTTT3= R: 5=CCTACAAAACCACTCRAAACTA3= F: 5=GGGTTGGTTTAGTTTYGTTAAGTTT3= R: 5=TAAAAATTTCCCAAAAACCTCTCC3= F: 5=TTTGGTTAGTAGTTTTIGGAGAAGA3= R: 5=CCCTATAAACCATAACTCCATAAACC3= F: 5=TGTTTAGTAGTTYGTTGTTYGGTTT3= R: 5=AACCCTCGCAAAATATCCAAC3= F: 5=GGTTTTGTGTTTTATTGYGGAGTG3= R: 5=CACCAATACAACCACATATCIATCAC3=

hMLH1-3= p16INK4a p14ARF MINT1 MINT2 MINT31 SFRP2 RASSF2A MGMT Reprimo 3OST2 HPP1 APC

Product size (restriction enzyme)

Temperature in Celsius (number of PCR cycles)

148 (HhaI)

59 (5), 57 (10), 55 (30)

160 (RsaI)

59 (5), 57 (10), 55 (30)

181 (TaqI)

58 (3), 56 (7), 54 (15), 52 (20)a

160 (TaqI)

60 (45)b

199 (TaqI)

55 (45)a

203 (TaqI)

60 (3), 58 (7), 56 (15), 54 (20)a

185 (HpyCH4IV)

58 (3), 56 (7), 54 (15), 52 (20)a

148 (HhaI)

59 (5), 57 (10), 55 (30)

131 (TaqI)

62 (5), 60 (10), 58 (30)

145 (BstUI)

53 (10), 50 (35)

138 (TaqI)

60 (15), 58 (30)

171 (TaqI)

60 (5), 58 (10), 56 (30)

137 (NruI)

58 (5), 56 (10), 54 (30)

156 (TaqI)

60 (45)

BASIC– ALIMENTARY TRACT

COBRA, combined bisulfite restriction assays; F, forward; R, reverse. conditions described previously by Rashid et al.35 bPCR conditions described previously by Shen et al.15 aPCR

(MSI-H) and are referred to as MSI cancers throughout the text. We then analyzed all non-MSI tumors with 5 additional dinucleotide repeat sequences (D5S107, D8S87, D17S261, D18S35, and D18S58) and 1 tetranucleotide repeat marker (MYCL1). MSI-low was defined as a shift in any of the dinucleotide and/or tetranucleotide markers; tumors that showed no allelic shifts were classified as microsatellite stable (MSS).31 We grouped 55 MSI-low and 152 MSS tumors together as non-MSI cancers in this study for comparative purposes because both have similar clinicopathologic and mutational features.

BRAF and KRAS Mutation Analysis Direct sequencing was performed to identify BRAF V600E mutation and KRAS codon 12/13 mutations. Primer sequences for BRAF and KRAS were as follows: BRAF-F (5=-TGCTTGCTCTGATAGGAAAATGA-3=), BRAF-R (5=TGGATCCAGACAACTGTTCAAA-3=), KRAS-F (5=-GCCTGCTGAAAATGACTGAA-3=), and KRAS-R (5=-AGAATGGTCCTGCACCAGTAA-3=) that generated fragment lengths of 165 and 167 base pairs, respectively. PCR products were purified using a QIAquick PCR purification kit (Qiagen, Valencia, CA) and directly sequenced on an ABI 3100-Avant DNA sequencer.

Statistical Analyses The methylation status of 14 epigenetic markers was analyzed as a categorical variable (methylated: methylation level ⱖ5%; unmethylated: methylation level ⬍5%). CRCs were divided into subgroups according to sporadic MSI/ Lynch syndrome or BRAF/KRAS mutation status, and the relationship of each epigenetic marker with various clinicopathologic parameters was evaluated using the ␹2 test. We used Kruskal–Wallis 1-way analysis of variance on rank sums to compare overall differences in the average number of methylated loci among subgroups classified by MSI status (sporadic MSI or Lynch syndrome) or BRAF/KRAS mutation status. Whenever the Kruskal–Wallis test indicated differences among these subgroups, further pair-wise comparisons for each of the subgroups divided by sporadic MSI/Lynch syndrome or BRAF/KRAS mutation status was performed using a nonparametric multiple comparison method using the Steel–Dwass test. To ascertain the relative risks for methylation, we calculated the odds ratio (OR) of each CRC subgroup according to methylation at each epigenetic marker. A 95% confidence interval (CI) was calculated for each OR. All reported P values are 2-sided, and P ⬍ .05 was considered statistically significant.

June 2008

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

A total of 487 colorectal tissues, including 243 CRCs, 208 C-Ns, and 36 N-Ns were investigated. All tumors were categorized into subgroups depending on their MSI status (sporadic or Lynch syndrome) and BRAF/KRAS mutations. We found that 36 of 243 (15%) CRCs were MSI and that 207 of 243 (85%) were non-MSI. Among the 36 MSI tumors, 15 (42%) were sporadic MSI CRCs, whereas 21 (58%) were from Lynch syndrome patients (which had been oversampled for this analysis). To highlight the differences between Lynch syndrome and sporadic MSI tumors, we categorized all CRCs into 3 subsets based on BRAF/KRAS mutation status as sporadic MSI, Lynch syndrome, and non-MSI cancers. We observed that 8% (20/243) of the CRCs harbored V600E BRAF mutations (“BRAF mutant”), 33% (80/243) had KRAS codon 12 or 13 mutations (“KRAS mutant”), and 59% (143/243) of CRCs lacked mutations in both BRAF and KRAS genes (“wild type”).

Table 2 illustrates detailed associations between the CRC groups based on MSI status (sporadic MSI and Lynch syndrome) and mutation spectrum (BRAF/KRAS). We observed that a significant proportion of sporadic MSI CRCs occurred in patients ⱖ65 years of age (80%), which was similar to the frequency in non-MSI patients (51%), but, as anticipated, the age was significantly lower in Lynch syndrome CRCs (5%, Lynch syndrome vs sporadic MSI or non-MSI; P ⬍ .0001). Sporadic MSI tumors were more frequent in females than males (sporadic MSI, 60%; Lynch syndrome, 19%; non-MSI, 35%). In addition, 92% of sporadic MSI and 69% of Lynch syndrome CRCs were proximally located, in contrast to 30% of non-MSI cancers. Older patients (ⱖ65 years at diagnosis) frequently harbored BRAF (65%) or KRAS (56%) mutations, although many lacked mutations in either gene (42%) (P ⫽ .0378 for either mutation vs younger patients). Female patients had significantly more BRAF mutations compared with males (BRAF mutant, 65%; KRAS mutant, 35%; wild type, 32%; P ⫽ .0147 for BRAF mutant vs KRAS mutant, P ⫽ .0033 for BRAF mutant vs wild-type). BRAF and KRAS mutant CRCs were significantly more common in the proximal colon,

Table 2. Associations Among Clinicopathologic Data, MSI Status, and Mutation Spectrum of the Ras Pathway Genes in all CRC Patients MSI status, % (n) MSI

Age, y ⱕ65 ⬍65 Sex Female Male Location Proximal Distal Not known Stage I–II III–IV Not known BRAF/KRAS mutation status BRAF mutant KRAS mutant Both wild type

BRAF/KRAS mutation status, % (n)

Sporadic (n ⫽ 15)

Lynch syndrome (n ⫽ 21)

Non-MSI (n ⫽ 207)

BRAF mutant (n ⫽ 20)

KRAS mutant (n ⫽ 80)

Wild type (n ⫽ 143)

80 (12) 20 (3)

5 (1) 95 (20)

51 (105) 49 (102)

65 (13) 35 (7)

56 (45) 44 (35)

42 (60) 58 (83)

65 (13) 35 (7)

35 (28) 65 (52)

32 (45) 69 (98)

30 (60) 70 (143) (4)

⬍.0001a .0283b ⬍.0001c .0117a .0553b .1343c .0912a ⬍.0001b .0079c

65 (11) 35 (6) (3)

46 (36) 54 (42) (2)

26 (36) 74 (101) (6)

.4782d .0523e .0403f .0147d .0033e .5899f .1507d .0013e .0047f

60 (9) 40 (6)

19 (4) 81 (17)

35 (73) 65 (134)

92 (12) 8 (1) (2)

69 (11) 31 (5) (5)

67 (8) 33 (4) (3)

67 (2) 33 (1) (18)

40 (83) 60 (123) (1)

1.000a .0717b .3558c

50 (10) 50 (10) (0)

39 (29) 61 (46) (5)

43 (54) 57 (72) (17)

.3600d .5498e .5595f

67 (10) 0 (0) 33 (5)

0 (0) 33 (7) 67 (14)

5 (10) 35 (73) 60 (124)

⬍.0001a ⬍.0001b .5546c

P value

P value

NOTE. All P values were calculated by the ␹2 test. MSI, microsatellite instability; BRAF mutant, CRCs with BRAF V600E mutation; KRAS mutant, CRCs with KRAS mutations; wild type, CRCs with neither BRAF nor KRAS mutations. aP values were calculated between sporadic MSI v Lynch syndrome. bP values were calculated between sporadic MSI v non-MSI cancers. cP values were calculated between Lynch syndrome v non-MSI. dP values were calculated between BRAF mutant v KRAS mutant. eP values were calculated between BRAF mutant v wild type. fP values were calculated between KRAS mutant v wild type.

BASIC– ALIMENTARY TRACT

Results V600E BRAF Mutations Are Frequently Present in Sporadic MSI CRCs, Whereas KRAS Mutations Are Exclusively Observed in Lynch Syndrome and Non-MSI CRCs

1953

1954

NAGASAKA ET AL

compared with wild-type CRCs (BRAF mutant, 65%; KRAS mutant, 46%; wild-type, 26%; P ⫽ .0013 for BRAF mutant vs wild-type, P ⫽ .0047 for KRAS mutant vs wild type). Importantly, BRAF mutations were detected at high frequencies in sporadic MSI tumors (67%), were seldom present in non-MSI tumors (5%), and did not occur in Lynch syndrome patients (0%). On the other hand, KRAS mutations were never present in sporadic MSI cancers (0%), whereas 33% of Lynch syndrome and 35% of nonMSI CRC exhibited KRAS mutations.

Aberrant DNA Hypermethylation Is Rare in Normal Colon but Is Frequently Observed in CRC We investigated the methylation status at 14 methylation-related loci using quantitative combined bisulfite restriction assays in the total collection of 487 colorectal tissues. We observed that 97% of CRCs (236/ 243), 46% of C-Ns (96/208), and 19% of N-Ns (7/36)

GASTROENTEROLOGY Vol. 134, No. 7

showed evidence for methylation at ⱖ1 markers (Figure 1A). Methylation levels in C-N and N-N tissues were quite low (⬍5% methylation) in comparison to that observed in CRCs, suggesting the cancer specificity of various methylation markers (Figure 1B). Furthermore, the proportion of tissues with methylation at ⱖ2 loci was significantly higher in CRCs (90%, 219/243) than in C-Ns (14%, 29/208) or N-Ns (3%, 1/36; P ⬍ .0001). As shown in Table 3, the overall frequency of promoter hypermethylation at each gene/locus was highly variable: SFRP2 (63%), RASSF2A (68%), MGMT (24%), Reprimo (24%), 3OST2 (82%), HPP1 (69%), APC (29%), hMLH1-5= region (22%), hMLH1-3= region (4%), p16INK4a (17%), p14ARF (10%), MINT1 (12%), MINT2 (21%), and MINT31 (26%).

A Distinct Subset of Markers Is Methylated in Sporadic MSI vs Lynch Syndrome CRCs We correlated the methylation status at individual epigenetic markers in the MSI tumors (Table 3).

BASIC– ALIMENTARY TRACT

Figure 1. Frequent hypermethylation at multiple loci in CRCs. (A) Distribution of all colorectal tissue specimens based on the number of methylated promoter loci. CRC denotes colorectal cancer; C-N denotes corresponding normal colonic epithelium; N-N denotes normal colonic epithelium without neoplasia at colonoscopy. (B) The Figure illustrates COBRA data for 7 canonical CIMP-related markers (hMLH1-5= region, hMLH1-3= region, p16INK4a, p14ARF, MINT1, MINT2, and MINT31) and the 7 new markers (SFRP2, RASSF2A, MGMT, Reprimo, 3OST2, HPP1, and APC) in CRCs (T), corresponding normal colonic epithelium (N), and normal colonic epithelium without neoplasia on colonoscopy (NN). Mc indicates SssI methylase-treated control human genomic DNA. PCR conditions of p16, MINT1, MINT2, and MINT31 were described previously by Rashid et al35 and for that of p14 by Shen et al.15

June 2008

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

1955

Table 3. Frequency of DNA Methylation at Each Epigenetic Marker in CRCs and Its Association With MSI Status or Mutation Spectrum of the BRAF/KRAS Genes MSI status, % (n)

Epigenetic marker hMLH1-5

Total, % (n), n ⫽ 243 22 (54)

Sporadic (n ⫽ 15) 60 (9)

BRAF/KRAS mutation status, % (n)

Lynch syndrome (n ⫽ 21) 29 (6)

Non-MSI (n ⫽ 207) 19 (39)

hMLH1-3

4 (9)

60 (9)

0 (0)

0 (0)

P16INK4a

17 (42)

53 (8)

5 (1)

16 (33)

P14ARF

10 (24)

33 (5)

14 (3)

8 (16)

MINT1

12 (30)

40 (6)

24 (5)

9 (19)

MINT2

21 (52)

53 (8)

14 (3)

20 (41)

MINT31

26 (62)

60 (9)

62 (13)

19 (40)

SFRP2

63 (154)

67 (10)

76 (16)

62 (128)

RASSF2A

68 (164)

87 (13)

62 (13)

67 (138)

MGMT

24 (59)

40 (6)

10 (2)

25 (51)

Reprimo

24 (57)

53 (8)

19 (4)

22 (45)

3OST2

82 (198)

93 (14)

76 (16)

81 (168)

HPP1

69 (168)

87 (13)

67 (14)

68 (141)

APC

29 (70)

27 (4)

48 (10)

27 (56)

P value .0593a .0002b .2857c ⬍.0001a ⬍.0001b NCc .0009a .0003b .1705c .1753a .0011b .3003c .2985a .0003b .0374c .0122a .0025b .5413c .9080a .0002b ⬍.0001c .5294a .7095b .1938c .1020a .1088b .6602c .0301a .1884b .1182c .0314a .0056b .7748c .1736a .2362b .5825c .1719a .1323b .8921c .2036a .9740b .0477c

BRAF mutant (n ⫽ 20)

KRAS mutant (n ⫽ 80)

Wild-type (n ⫽ 143)

65 (13)

19 (15)

18 (26)

0 (0)

1 (2)

55 (11)

21 (17)

10 (14)

20 (4)

13 (10)

7 (10)

40 (8)

18 (14)

6 (8)

80 (16)

26 (21)

11 (15)

60 (12)

21 (17)

23 (33)

55 (11)

80 (64)

55 (79)

75 (15)

83 (66)

58 (83)

30 (6)

40 (32)

15 (21)

60 (12)

34 (27)

13 (18)

80 (16)

94 (75)

75 (107)

75 (15)

80 (64)

62 (89)

25 (5)

29 (23)

29 (42)

35 (7)

P value ⬍.0001d ⬍.0001e .9163f ⬍.0001d ⬍.0001b .2880f .0026d ⬍.0001e .0177f .3873d .0518e .1675f .0298d ⬍.0001e .0042f ⬍.0001d ⬍.0001e .0022f .0006d .0005e .7537f .0209d .9836e .0002f .4444d .1469e .0002f .4099d .0844e ⬍.0001f .0313d ⬍.0001e .0002f .0546d .6145e .0005f .6234d .2660e .0061f .7383d .6861e .9221f

NOTE. All P values were calculated by the ␹2 test. NC, denotes “could not calculate.” aP values were calculated between sporadic MSI v Lynch syndrome. bP values were calculated between sporadic MSI v non-MSI cancers. cP values were calculated between Lynch syndrome v non-MSI. dP values were calculated between BRAF mutant v KRAS mutant. eP values were calculated between BRAF mutant v wild type. fP values were calculated between KRAS mutant v wild type.

When methylation features were compared between sporadic MSI and Lynch syndrome tumors, as expected, sporadic MSI tumors were more frequently methylated at most of the markers, the exceptions

being MINT31, SFRP2, and APC. Interestingly, aberrant methylation at these 3 loci was more frequent in Lynch syndrome tumors. MINT31, which is one of the canonical CIMP markers, was methylated at equal frequen-

BASIC– ALIMENTARY TRACT

MSI

NAGASAKA ET AL

GASTROENTEROLOGY Vol. 134, No. 7

cies in both Lynch syndrome (62%) and sporadic MSI CRCs (60%) as in sporadic MSI (60%). Of potential importance for understanding the genesis of tumors in Lynch syndrome, methylation at the APC promoter was more frequently observed in Lynch syndrome CRCs (48%) in comparison with both sporadic MSI (27%) and non-MSI (27%) tumors.

Hypermethylation Frequency at Individual Markers Is Related to BRAF and KRAS Mutational Status in CRC

A

We estimated the OR for methylation at each marker among subgroups of CRCs segregated by sporadic MSI/Lynch syndrome and BRAF/KRAS mutation status (Figure 2 and see Supplemental Table 1 online at www.gastrojournal.org). When ORs were calculated for sporadic MSI vs Lynch syndrome CRCs, we observed that methylation of all 7 canonical CIMP markers, plus Reprimo, showed a positive risk for methylation in sporadic MSI tumors (Figure 2A). Contrariwise, methylation of MINT31 was significantly associated with Lynch syndrome cancers (OR, 5.74; 95% CI: 2.29 –15.2). None of the other noncanonical epigenetic markers showed any risk associations based on MSI and/or Lynch syndrome status. We then examined ORs for methylation at each marker among the subgroups of CRCs categorized by BRAF/ KRAS mutation status (Figure 2B). The ORs of the 6 canonical markers (hMLH1-5= region, hMLH1-3= region, p16INK4a, MINT1, MINT2, and MINT31), together with Reprimo, were significantly higher in BRAF mutant CRCs, whereas 5 of the 7 new markers (SFRP2, RASSF2A, MGMT, 3OST2, and HPP1) showed significantly higher ORs in KRAS mutant CRCs. Although the OR for p14ARF was the highest for BRAF mutant cancers, it was not

B 103

Sporadic MSI

102

Odds Ratio

10 1

102

1

102

Lynch Syndrome

10

Odds Ratio

1

10-1 10-2

KRAS

10 1 10-1 10-4 10

CIMP-markers

Additional-markers

CIMP-markers

APC

HPP1

3OST2

MGMT

Reprimo

10-2

SFRP2

10-1

RASSF2A

APC

HPP1

3OST2

MGMT

Reprimo

SFRP2

RASSF2A

MINT31

MINT1

MINT2

p14ARF

p16INK4a

hMLH1-5

hMLH1-3

10-1

1

MINT31

1

No Mutant

MINT1

Odds Ratio

Non-MSI

p14ARF

10

MINT2

10-4

p16INK4a

Odds Ratio

10

10-1

10-1

10-4

BRAF

102

hMLH1-5

Odds Ratio

107

Odds Ratio

BASIC– ALIMENTARY TRACT

We investigated the relationship between methylation frequencies at each of the 14 epigenetic markers and the presence or absence of BRAF/KRAS mutations in the total collection of CRCs (Table 3). We observed that the 6 classical CIMP-related markers (hMLH1-5= region, hMLH1-3= region, p16INK4a, MINT1, MINT2, and MINT31) along with Reprimo were significantly more frequently methylated in BRAF mutant CRCs compared with the other subgroups. However, we found that methylation frequencies at the other 5 markers (SFRP2, RASSF2A, MGMT, 3OST2, and HPP1) were almost the same or relatively higher in CRCs harboring KRAS mutations. At the same time, APC methylation frequencies were comparable in all 3 subsets (25% in BRAF mutants, 29% in KRAS mutants, and 29% in wild types).

Epigenetic Alterations Can Be Interpreted in the Context of the Mutational Spectrum of the KRAS-BRAF Genes

hMLH1-3

1956

Additional-markers

Figure 2. The odds ratio (OR) for methylation at each epigenetic marker in subgroups of CRCs categorized by MSI status (sporadic or Lynch syndrome) (panel A) or BRAF/KRAS mutational status (panel B). (A) The upper, middle, and bottom panels illustrate the ORs for sporadic MSI tumors (MSI), Lynch syndrome CRCs, and non-MSI CRCs, respectively. (B) The upper, middle, and bottom panels illustrate the ORs for BRAF mutant CRCs (BRAF), KRAS mutant CRCs (KRAS), and wild-type (both wild type) CRCs, respectively. The vertical bars depict the 95% CIs for this ratio. An odds ratio (OR) ⬎1.0 for a given marker is represented by a white square and suggests a positive association with that subgroup (⬎1.0), whereas an OR ⬍1.0 represents a negative risk and is shown as black squares. An OR for an epigenetic marker that does not show any positive or negative associations is shown as a gray square. All ORs are presented in a log scale.

June 2008

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

Aberrant DNA Methylation Is More Frequent in Lynch Syndrome CRCs Than in Non-MSI CRCs Both Lynch syndrome CRCs and tumors with methylated hMLH1 promoters have the MSI phenotype, but they arise through different pathways. To understand better the differences between these types of CRCs, we determined the average numbers of methylated loci in each subgroup of CRCs (Figure 3A and Table 4). When methylation data were utilized using all 14 epigenetic markers, the average number of methylated loci was highest in sporadic MSI CRCs (8.4; 95% CI: 6.0 –10.2), followed by Lynch syndrome CRCs (5.1; 95% CI: 4.1– 6.2), and was least in non-MSI CRCs (4.4;

2

0.9

1

4 3

1.2

2

0.7

1

3 2 1 Non-MSI

Sporadic MSI 7 6 5

P < .0001 4.4

4.3 3.2

4 3 2 1 0

No Mutant

KRAS

BRAF

No Mutant

KRAS

BRAF

3.6

4

Non-MSI

Lynch syndrome

P < .0001 3.5

0

0

3. 7

No Mutant

3.9

5

6 5

P = .0177 4.8

Additional-markers

CIMP-markers

7

No. of Methylated Loci

5. 5

No. of Methylated Loci

P < .0001

8.0

5

0 Sporadic MSI

Non-MSI

Lynch syndrome

Sporadic MSI No. of Methylated Loci

1.5

All Loci

15

10

3

0

0

B

4

6

Lynch syndrome

4.4 5

3.6

5

7

KRAS

5.1

6

P < .0001

BRAF

P < .0001

7

No. of Methylated Loci

8.4

10

Additional-markers

CIMP-markers No. of Methylated Loci

All Loci

15 No. of Methylated Loci

A

95% CI: 4.1– 4.8). Not surprisingly, when the data were analyzed using only the canonical CIMP markers (hMLH1-5= region, hMLH1-3= region, p16INK4a, p14ARF, MINT1, MINT2, and MINT31), the average number of methylated loci was significantly higher in sporadic MSI CRCs (3.6; 95% CI: 2.3– 4.9), compared with Lynch syndrome (1.5; 95% CI: 1.0 –2.0), or non-MSI subgroups (0.9; 95% CI: 0.7–1.0; P ⬍ .0001). However, interestingly, when the data were analyzed only using the 7 additional methylation markers, the differences in average number of methylated loci between Lynch syndrome and non-MSI tumors became much smaller (sporadic MSI [4.8; 95% CI: 4.1–5.5], Lynch syndrome [3.7; 95% CI: 3.0 – 4.4], non-MSI [3.6; 95% CI: 3.3–3.8]). Furthermore, the average numbers of methylated loci were consistently higher for Lynch syndrome CRCs compared with non-MSI CRCs (Table 4 and Figure 3A). More specifically, Lynch syndrome CRCs showed significantly higher methylation when data were analyzed from 7 CIMP canonical markers (1.5 vs 0.9, respectively; P ⫽ .0230). Although Lynch syndrome cancers were more frequently methylated than non-MSI cancers, these data did not reach significance when comparisons were drawn from 14 unselected epigenetic markers (5.1 vs 4.4, respec-

Figure 3. Average numbers of methylated loci in various subgroups of CRCs categorized by MSI status (panel A) or BRAF/KRAS mutation status (panel B). The average number of methylated loci in each subset was calculated by all 14 markers (all loci), 7 canonical CIMP markers (CIMP-related), or the 7 additional markers (additional markers). In the box plot diagrams, the horizontal line within each box represents the median; the limits of each box are the interquartile ranges, the whiskers are the maximum and minimum values, and the blue cross within each box depicts the mean value. The numbers above each box denotes the mean number of methylated loci. The P values above the square panels were based on Kruskal–Wallis 1-way analysis of variance on ranks and represent the statistical differences in average methylation among all 3 subsets (sporadic MSI, Lynch syndrome, and non-MSI) of CRCs. Statistical differences among any 2 individual groups are shown as pair-wise comparisons in Table 4.

BASIC– ALIMENTARY TRACT

significantly different compared with the other 2 subgroups. Similarly, the OR for APC methylation was similar among various subsets (OR for BRAF mutant, 0.81; for KRAS mutant, 1.00; for wild type, 1.07). This analysis permitted us to categorize the epigenetic markers into 3 distinct subsets: BRAF-related markers (hMLH1-5= region, hMLH1-3= region, p16INK4a, MINT1, MINT2, MINT31, and Reprimo), KRAS-related markers (SFRP2, RASSF2A, MGMT, 3OST2, and HPP1), and markers independent of the BRAF/ KRAS mutations (p14ARF and APC).

1957

1958

NAGASAKA ET AL

GASTROENTEROLOGY Vol. 134, No. 7

Table 4. Pair-Wise Comparisons of Methylation Levels for Various Epigenetic Markers With Different Subgroups of CRCs Segregated Based on Their MSI Status and BRAF/KRAS Mutation Status Subset of markers All markers

CIMP markers

Additional markers

MSI status Pair-wise comparison

P value

BRAF/KRAS mutation status Pair-wise comparison

P value

Sporadic MSI v Lynch syndrome Sporadic MSI v non-MSI Lynch syndrome v non-MSI Sporadic MSI v Lynch syndrome Sporadic MSI v non-MSI Lynch syndrome v non-MSI Sporadic MSI v Lynch syndrome Sporadic MSI v non-MSI Lynch syndrome v non-MSI

.0101 ⬍.0001 .2782 .0123 ⬍.0001 .0230 .0509 .0124 .9977

BRAF v KRAS BRAF v wild type KRAS v wild type BRAF v KRAS BRAF v wild type KRAS v wild type BRAF v KRAS BRAF v wild type KRAS v wild type

.0042 ⬍.0001 ⬍.0001 ⬍.0001 ⬍.0001 .0353 .8610 .00883 ⬍.0001

NOTE. P values were based on Steel–Dwass test.

tively; P ⫽ .2782) or the 7 additional markers (3.7 vs 3.6, respectively; P ⫽ .9977).

Comparable Methylation Frequencies Occur in BRAF and KRAS Mutant CRCs Analyzed Using the Additional Epigenetic Markers

BASIC– ALIMENTARY TRACT

We then determined whether the average numbers of methylated loci were influenced by the presence of BRAF and KRAS mutations in CRC (Figure 3B and Table 4). When we used data from all 14 markers, we observed that the average number of methylated loci was highest in BRAF mutated tumors (8.0; 95% CI: 6.5–9. 5), followed by KRAS mutants (5.5; 95% CI: 5.0 – 6.0) and the lowest in CRCs with wild-type BRAF and KRAS genes (3.9; 95% CI: 3.5– 4.3; P ⬍ .0001). Similar findings were observed when data were analyzed from the 7 canonical CIMP markers alone (BRAF mutant [3.55; 95% CI: 2.7– 4.4], KRAS mutant [1.2; 95% CI: 0.9 –1.5], wild type [0.7; 95% CI: 0.6 – 0. 9]; P ⬍ .0001). Interestingly, when the data were analyzed using the 7 additional markers, the average number of methylated loci in KRAS mutated cancers was no different than that in BRAF mutant CRCs (BRAF mutant [4.4; 95% CI: 3.7– 5.1], KRAS mutant [4.3; 95% CI: 4.0 – 4.6]; P ⫽ .8610), and the average number of methylated loci in both of these subgroups was significantly higher than in the tumors lacking mutations in both BRAF and KRAS (3.2; 95% CI: 2.9 –3.5; BRAF vs wild type, P ⬍ .01; KRAS vs wild type, P ⬍ .0001).

Discussion This study investigates the relationship between mutational activation in the RAS-RAF signaling pathway and global hypermethylation using a panel of epigenetic markers in a collection of 487 colorectal tissues. More specifically, we asked whether CIMP in the colon is exclusively correlated with BRAF mutations or whether it may also associate with mutant KRAS because both genes are members of the same signaling pathway. Additionally, we asked whether CIMP is primarily a disease of older individuals with sporadic MSI CRCs or whether

Lynch syndrome CRCs may also have molecular features consistent with CIMP. We have provided evidence that aberrant hypermethylation of various tumor suppressor genes and related loci not only associate with mutant BRAF but also with mutant KRAS and that activation of the KRAS-BRAF pathway induces aberrant promoter methylation in multiple genes. Additionally, we found that Lynch syndrome CRCs have frequent methylation, challenging the supposition that CIMP is exclusively a molecular characteristic of sporadic CRCs. Although the CIMP concept was first proposed in CRC almost a decade ago,2 only recent evidence has supported its existence in a specific subset of sporadic CRCs.3,4,8,9 In spite of this, the molecular basis of CIMP remains unclear and is a matter of active investigation. Currently, CIMP CRCs are characterized using a panel of markers that were selected, in part, through their association with the V600E BRAF mutation.11–13 Although the canonical CIMP-related markers are highly specific, this panel may not be completely adequate because there is no consensus on the frequency of CIMP in CRC.8,9 Part of the problem may be attributed to differences in methodologies for measuring methylation in each laboratory, but central to this issue is a lack of consensus criteria and definitions for CIMP in CRC.8,9,14,15 Moreover, some studies have reported data in which a subset of CRCs have intermediate levels of aberrant DNA methylation defined as CIMP-low because they failed to meet the more restrictive criteria for CIMP.14,32–35 This has led others to challenge the methylator phenotype as a discrete “pathway” in colorectal carcinogenesis. Herein, we present data that clearly suggest that CIMPassociated aberrant methylation observed in CRCs may not be limited to those with V600E BRAF mutations. Rather, CIMP determination appears to be dependent on the choice of methylation markers. When data are analyzed using canonical CIMP-related markers, one can identify a distinct cluster of CRCs that are strongly associated with mutant BRAF.14,33 However, this tight clustering of CRC disappears when additional epigenetic markers are analyzed. We demonstrate that, when data

are evaluated using canonical markers, BRAF mutant CRCs have the highest average number of methylated loci. However, when additional methylation markers are analyzed, both BRAF and KRAS mutated cancers show similar degrees of methylation, and methylation levels in these 2 subgroups are significantly higher than in CRCs that are wild type for these genes. In comparison with the classical CIMP markers that strongly associate with mutant BRAF, data analysis with noncanonical methylation markers in our study resulted in loss of the typical bimodal distribution observed for CIMP-positive and -negative tumors. However, these data are biologically relevant because a significant amount of aberrant methylation was associated with mutant KRAS. Additionally, when data were analyzed from only nonCIMP markers, KRAS mutated and wild-type CRCs were statistically distinguishable from each other (Table 4). Similar suggestions have been made in previous studies.11,36 It is difficult to compare directly our panel of noncanonical CIMP markers with the markers investigated in the previous studies.11,36 However, a common feature among these studies is that a significant amount of methylation positively associates with mutant KRAS in the colon. Because KRAS and BRAF belong to the same growth signaling pathway, these data argue that activating mutations in either of these genes may have equivalent effects in mediating aberrant DNA methylation. This brings the RAS/RAF story full circle because it was initially found that KRAS mutations were responsible for up-regulating DNA methylation and a methylator phenotype in vitro.20 In support of this, a recent study has clearly shown that Ras-mediated epigenetic silencing occurs through a specific, but complex, pathway involving components that are essential for maintaining a fully transformed phenotype in a fibroblast cell line.37 It has been proposed that CIMP CRCs that harbor BRAF mutations and demonstrate MSI may originate through a unique pathway that includes the progression of sessile serrated polyps to MSI CRCs.38 It has been suggested that V600E BRAF mutations are present in sessile serrated polyps and serrated aberrant crypt foci, whereas KRAS mutations are more highly associated with nonserrated polyps and aberrant crypt foci.32,39 Interestingly, 90% of aberrant crypt foci with BRAF mutations was found to be MSS.39 Approximately 70% of sporadic MSI CRCs exhibits BRAF mutations, and a majority of sporadic MSI CRC is caused by extensive hypermethylation of the hMLH1 gene.17,32 Our present findings and data and that from others suggest that aberrant DNA methylation may be induced by up-regulation of the RAS-RAF pathway37 and that genetic alteration in BRAF or KRAS might be an earlier event that precedes aberrant DNA methylation. In this study, we noticed a positive association of RASSF2A methylation with mutant KRAS, which contradicts an earlier report.23 However, the study by Hesson

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

1959

et al23 involved a small subset of tissues (8 adenomas and 33 carcinomas). A subsequent study analyzed a group of 140 CRCs and reported that 67% of CRCs with RASSF2 methylation displayed BRAF or KRAS mutations (P ⫽ .0009) and showed that inactivation of RASSF2 enhanced KRAS-induced oncogenic transformation.25 Our present data are in agreement with this recent study, and we feel confident that our data are reliable and that RASSF2A methylation positively associates with mutant KRAS in the colon. CIMP predominantly occurs in older individuals with sporadic MSI CRCs, but it has not been rigorously investigated in Lynch syndrome.32,40 In this study, we observed that sporadic MSI CRCs had the highest degree of aberrant methylation regardless of the epigenetic markers analyzed.32 However, unexpectedly, we observed that Lynch syndrome patients, who were much younger, had a higher degree of methylation than that of non-MSI CRCs, particularly when data were analyzed from 7 canonical CIMP markers (P ⫽ .023). It is possible that the higher degree of methylation observed in Lynch syndrome may be due, in part, to methylation events associated with the frequent KRAS mutations. In the past decade, the efforts of several laboratories have been focused on the molecular mechanisms of aberrant DNA methylation, but our current understanding into these processes is limited. Although this study does not provide a conclusive explanation for the specific processes that control DNA methylation, our data provide indirect evidence that highlight the importance of KRAS-associated methylation events. We also found that aberrant DNA methylation is a much more common event in Lynch syndrome patients than was previously presumed. We speculate that genetic alterations in the BRAF and KRAS oncogenes are an early event in the evolution of a methylator phenotype and that regulation of aberrant DNA methylation may be located downstream in the RAS-RAF signaling pathway. Finally, considering the paucity of information on the causes of CIMP, in future studies, it may be prudent to interpret aberrant DNA methylation in the context of mutations in both BRAF and KRAS genes.

Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi:10.1053/j.gastro. 2008.02.094 References 1. Herman JG, Baylin SB. Gene silencing in cancer in association with promoter hypermethylation. N Engl J Med 2003;349:2042– 2054. 2. Toyota M, Ahuja N, Ohe-Toyota M, et al. CpG island methylator phenotype in colorectal cancer. Proc Natl Acad Sci U S A 1999; 96:8681– 8686.

BASIC– ALIMENTARY TRACT

June 2008

1960

NAGASAKA ET AL

BASIC– ALIMENTARY TRACT

3. Issa JP, Shen L, Toyota M. CIMP, at last. Gastroenterology 2005; 129:1121–1124. 4. Goel A, Nagasaka T, Arnold CN, et al. The CpG island methylator phenotype and chromosomal instability are inversely correlated in sporadic colorectal cancer. Gastroenterology 2007;132:127–138. 5. Eads CA, Lord RV, Wickramasinghe K, et al. Epigenetic patterns in the progression of esophageal adenocarcinoma. Cancer Res 2001;61:3410 –3418. 6. Esteller M, Tortola S, Toyota M, et al. Hypermethylation-associated inactivation of p14(ARF) is independent of p16(INK4a) methylation and p53 mutational status. Cancer Res 2000;60:129 –133. 7. Yamashita K, Dai T, Dai Y, et al. Genetics supersedes epigenetics in colon cancer phenotype. Cancer Cell 2003;4:121–131. 8. Samowitz WS, Albertsen H, Herrick J, et al. Evaluation of a large, population-based sample supports a CpG island methylator phenotype in colon cancer. Gastroenterology 2005;129:837– 845. 9. Ogino S, Kawasaki T, Kirkner GJ, et al. Evaluation of markers for CpG island methylator phenotype (CIMP) in colorectal cancer by a large population-based sample. J Mol Diagn 2007;9:305–314. 10. Davies H, Bignell GR, Cox C, et al. Mutations of the BRAF gene in human cancer. Nature 2002;417:949 –954. 11. Weisenberger DJ, Siegmund KD, Campan M, et al. CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat Genet 2006;38:787–793. 12. Toyota M, Ohe-Toyota M, Ahuja N, et al. Distinct genetic profiles in colorectal tumors with or without the CpG island methylator phenotype. Proc Natl Acad Sci U S A 2000;97:710 –715. 13. Toyota M, Issa JP. CpG island methylator phenotypes in aging and cancer. Semin Cancer Biol 1999;9:349 –357. 14. Samowitz WS, Albertsen H, Sweeney C, et al. Association of smoking, CpG island methylator phenotype, and V600E BRAF mutations in colon cancer. J Natl Cancer Inst 2006;98:1731–1738. 15. Shen L, Kondo Y, Hamilton SR, et al. P14 methylation in human colon cancer is associated with microsatellite instability and wild-type p53. Gastroenterology 2003;124:626 – 633. 16. Slattery ML, Curtin K, Sweeney C, et al. Diet and lifestyle factor associations with CpG island methylator phenotype and BRAF mutations in colon cancer. Int J Cancer 2007;120:656 – 663. 17. Nagasaka T, Sasamoto H, Notohara K, et al. Colorectal cancer with mutation in BRAF, KRAS, and wild-type with respect to both oncogenes showing different patterns of DNA methylation. J Clin Oncol 2004;22:4584 – 4594. 18. Iacopetta B, Grieu F, Li W, et al. APC gene methylation is inversely correlated with features of the CpG island methylator phenotype in colorectal cancer. Int J Cancer 2006;119:2272–2278. 19. Kim JS, Lee C, Foxworth A, et al. B-Raf is dispensable for K-Rasmediated oncogenesis in human cancer cells. Cancer Res 2004; 64:1932–1937. 20. Ordway JM, Williams K, Curran T. Transcription repression in oncogenic transformation: common targets of epigenetic repression in cells transformed by Fos, Ras or Dnmt1. Oncogene 2004; 23:3737–3748. 21. Xiong Z, Laird PW. COBRA: a sensitive and quantitative DNA methylation assay. Nucleic Acids Res 1997;25:2532–2534. 22. Suzuki H, Watkins DN, Jair KW, et al. Epigenetic inactivation of SFRP genes allows constitutive WNT signaling in colorectal cancer. Nat Genet 2004;36:417– 422. 23. Hesson LB, Wilson R, Morton D, et al. CpG island promoter hypermethylation of a novel Ras-effector gene RASSF2A is an early event in colon carcinogenesis and correlates inversely with K-ras mutations. Oncogene 2005;24:3987–3994. 24. Takahashi T, Shigematsu H, Shivapurkar N, et al. Aberrant promoter methylation of multiple genes during multistep pathogenesis of colorectal cancers. Int J Cancer 2006;118:924 –931.

GASTROENTEROLOGY Vol. 134, No. 7

25. Akino K, Toyota M, Suzuki H, et al. The Ras effector RASSF2 is a novel tumor-suppressor gene in human colorectal cancer. Gastroenterology 2005;129:156 –169. 26. Esteller M, Sparks A, Toyota M, et al. Analysis of adenomatous polyposis coli promoter hypermethylation in human cancer. Cancer Res 2000;60:4366 – 4371. 27. Young J, Biden KG, Simms LA, et al. HPP1: a transmembrane protein-encoding gene commonly methylated in colorectal polyps and cancers. Proc Natl Acad Sci U S A 2001;98:265–270. 28. Miyamoto K, Asada K, Fukutomi T, et al. Methylation-associated silencing of heparan sulfate D-glucosaminyl 3-O-sulfotransferase-2 (3-OST-2) in human breast, colon, lung and pancreatic cancers. Oncogene 2003;22:274 –280. 29. Nagasaka T, Sharp GB, Notohara K, et al. Hypermethylation of O6-methylguanine-DNA methyltransferase promoter may predict nonrecurrence after chemotherapy in colorectal cancer cases. Clin Cancer Res 2003;9:5306 –5312. 30. Boland CR, Thibodeau SN, Hamilton SR, et al. A National Cancer Institute Workshop on Microsatellite Instability for cancer detection and familial predisposition: development of international criteria for the determination of microsatellite instability in colorectal cancer. Cancer Res 1998;58:5248 –5257. 31. Kambara T, Matsubara N, Nakagawa H, et al. High frequency of low-level microsatellite instability in early colorectal cancer. Cancer Res 2001;61:7743–7746. 32. Kambara T, Simms LA, Whitehall VL, et al. BRAF mutation is associated with DNA methylation in serrated polyps and cancers of the colorectum. Gut 2004;53:1137–1144. 33. Ogino S, Kawasaki T, Kirkner GJ, et al. CpG island methylator phenotype-low (CIMP-low) in colorectal cancer: possible associations with male sex and KRAS mutations. J Mol Diagn 2006;8: 582–588. 34. An C, Choi IS, Yao JC, et al. Prognostic significance of CpG island methylator phenotype and microsatellite instability in gastric carcinoma. Clin Cancer Res 2005;11:656 – 663. 35. Rashid A, Shen L, Morris JS, et al. CpG island methylation in colorectal adenomas. Am J Pathol 2001;159:1129 –1135. 36. Shen L, Toyota M, Kondo Y, et al. Integrated genetic and epigenetic analysis identifies three different subclasses of colon cancer. Proc Natl Acad Sci U S A 2007;104:18654 –18659. 37. Gazin C, Wajapeyee N, Gobeil S, et al. An elaborate pathway required for Ras-mediated epigenetic silencing. Nature 2007; 449:1073–1077. 38. Jass JR, Whitehall VL, Young J, et al. Emerging concepts in colorectal neoplasia. Gastroenterology 2002;123:862– 876. 39. Rosenberg DW, Yang S, Pleau DC, et al. Mutations in BRAF and KRAS differentially distinguish serrated versus non-serrated hyperplastic aberrant crypt foci in humans. Cancer Res 2007;67: 3551–3554. 40. Yamamoto H, Min Y, Itoh F, et al. Differential involvement of the hypermethylator phenotype in hereditary and sporadic colorectal cancers with high-frequency microsatellite instability. Genes Chromosomes Cancer 2002;33:322–325.

Received October 26, 2007. Accepted February 28, 2008. Address requests for reprints to: Ajay Goel, PhD, or C. Richard Boland, MD, Baylor University Medical Center, 3500 Gaston Avenue, Gastrointestinal Cancer Research Laboratory, Suite H-250, Dallas, Texas 75246. e-mail: [email protected] or Rickbo@baylorhealth. edu; fax: (214) 818-9292. Supported by grants R01 CA72851 and R01 CA98572 from the National Cancer Institute, National Institutes of Health (to C.R.B.), and funds from the Baylor Research Institute. Conflicts of interest: None of the authors have any conflicts of interest.

June 2008

RAS-RAF SIGNALING AND CIMP IN COLON CANCER

1960.e1

Supplemental Table 1. The Odds Ratios for Methylation for Each Epigenetic Marker in Various Subgroups of CRCs MSI status MSI

BRAF/KRAS mutation status

Marker

Sporadic MSI (n ⫽ 15)

Lynch syndrome (n ⫽ 21)

Non-MSI (n ⫽ 207)

BRAF mutant (n ⫽ 20)

KRAS mutant (n ⫽ 80)

No mutant (n ⫽ 143)

hMLH1-5 hMLH1-3 p16INK4a p14ARF MINT1 MINT2 MINT31 SFRP2 RASSF2A MGMT Reprimo 3OST2 HPP1 APC

6.10 (2.09⫺19.0) 2.790000 (NC) 6.52 (2.21⫺19.8) 5.50 (1.58⫺17.3) 5.67 (1.77⫺17.1) 4.78 (1.63⫺14.3) 4.95 (1.71⫺15.4) 1.17 (0.40⫺3.85) 3.31 (0.89⫺21.5) 2.20 (0.71⫺6.39) 4.17 (1.43⫺12.5) 3.35 (0.65⫺61.4) 3.06 (0.82⫺19.9) 0.89 (0.24⫺2.72)

1.45 (0.50⫺3.78) 0.0004 (NC⫺2.53) 0.22 (0.01⫺1.18) 1.60 (0.35⫺5.22) 2.46 (0.75⫺6.92) 0.59 (0.13⫺1.83) 5.74 (2.29⫺15.2) 1.95 (0.73⫺6.13) 0.76 (0.31⫺2.01) 0.30 (0.05⫺1.09) 0.75 (0.21⫺2.13) 0.70 (0.26⫺2.25) 0.88 (0.35⫺2.42) 2.45 (0.98⫺6.11)

0.33 (0.15⫺0.70) 0.000005 (NC⫺0.03) 0.57 (0.25⫺1.38) 0.29 (0.12⫺0.78) 0.23 (0.10⫺0.55) 0.56 (0.26⫺1.27) 0.15 (0.07⫺0.32) 0.62 (0.27⫺1.32) 0.77 (0.34⫺1.64) 1.14 (0.51⫺2.83) 0.56 (0.26⫺1.23) 0.86 (0.31⫺2.09) 0.71 (0.30⫺1.55) 0.58 (0.28⫺1.24)

8.24 (3.18⫺23.2) 59.5 (12.9⫺428) 7.57 (2.91⫺20.3) 2.54 (0.68⫺7.74) 6.09 (2.18⫺16.4) 20.8 (7.15⫺75.8) 5.19 (2.04⫺13.9) 0.68 (0.27⫺1.76) 1.49 (0.55⫺4.72) 1.37 (0.47⫺3.61) 5.93 (2.31⫺16.0) 0.90 (0.31⫺3.27) 1.37 (0.51⫺4.35) 0.81 (0.26⫺2.19)

0.73 (0.37⫺1.41) 0.0007 (NC⫺0.47) 1.49 (0.74⫺2.94) 1.52 (0.63⫺3.57) 1.95 (0.89⫺4.23) 1.52 (0.80⫺2.84) 0.71 (0.37⫺1.32) 3.24 (1.76⫺6.24) 3.13 (1.66⫺6.22) 3.36 (1.83⫺6.22) 2.26 (1.23⫺4.16) 4.88 (2.01⫺14.6) 2.27 (1.23⫺4.39) 1.00 (0.55⫺1.79)

0.57 (0.31⫺1.05) 0.19 (0.03⫺0.80) 0.28 (0.13⫺0.56) 0.46 (0.19⫺1.08) 0.21 (0.08⫺0.48) 0.20 (0.10⫺0.38) 0.73 (0.41⫺1.32) 0.41 (0.23⫺0.71) 0.32 (0.17⫺0.58) 0.28 (0.15⫺0.51) 0.22 (0.12⫺0.42) 0.29 (0.13⫺0.62) 0.44 (0.24⫺0.80) 1.07 (0.61⫺1.89)

NOTE. Values represent odds ratios and 95% CIs. Odds ratios were calculated for the odds of sporadic MSI (n ⫽ 15) vs others (n ⫽ 228), Lynch syndrome (n ⫽ 21) vs others (n ⫽ 222), and non-MSI (n ⫽ 207) vs others (n ⫽ 36); the odds of BRAF mutant (n ⫽ 20) vs others (n ⫽ 223), KRAS mutant (n ⫽ 80) vs others (n ⫽ 163), and wild type (n ⫽ 143) vs others (n ⫽ 100). NC, denotes “could not calculate.”